vfs_mount.c revision 146125
1/*-
2 * Copyright (c) 1999-2004 Poul-Henning Kamp
3 * Copyright (c) 1999 Michael Smith
4 * Copyright (c) 1989, 1993
5 *	The Regents of the University of California.  All rights reserved.
6 * (c) UNIX System Laboratories, Inc.
7 * All or some portions of this file are derived from material licensed
8 * to the University of California by American Telephone and Telegraph
9 * Co. or Unix System Laboratories, Inc. and are reproduced herein with
10 * the permission of UNIX System Laboratories, Inc.
11 *
12 * Redistribution and use in source and binary forms, with or without
13 * modification, are permitted provided that the following conditions
14 * are met:
15 * 1. Redistributions of source code must retain the above copyright
16 *    notice, this list of conditions and the following disclaimer.
17 * 2. Redistributions in binary form must reproduce the above copyright
18 *    notice, this list of conditions and the following disclaimer in the
19 *    documentation and/or other materials provided with the distribution.
20 * 4. Neither the name of the University nor the names of its contributors
21 *    may be used to endorse or promote products derived from this software
22 *    without specific prior written permission.
23 *
24 * THIS SOFTWARE IS PROVIDED BY THE AUTHOR AND CONTRIBUTORS ``AS IS'' AND
25 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
26 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
27 * ARE DISCLAIMED.  IN NO EVENT SHALL THE AUTHOR OR CONTRIBUTORS BE LIABLE
28 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
29 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
30 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
31 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
32 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
33 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
34 * SUCH DAMAGE.
35 */
36
37#include <sys/cdefs.h>
38__FBSDID("$FreeBSD: head/sys/kern/vfs_mount.c 146125 2005-05-12 01:49:51Z pjd $");
39
40#include <sys/param.h>
41#include <sys/conf.h>
42#include <sys/jail.h>
43#include <sys/kernel.h>
44#include <sys/libkern.h>
45#include <sys/mac.h>
46#include <sys/malloc.h>
47#include <sys/mount.h>
48#include <sys/mutex.h>
49#include <sys/namei.h>
50#include <sys/proc.h>
51#include <sys/filedesc.h>
52#include <sys/reboot.h>
53#include <sys/syscallsubr.h>
54#include <sys/sysproto.h>
55#include <sys/sx.h>
56#include <sys/sysctl.h>
57#include <sys/sysent.h>
58#include <sys/systm.h>
59#include <sys/vnode.h>
60
61#include <geom/geom.h>
62
63#include <machine/stdarg.h>
64
65#include "opt_rootdevname.h"
66#include "opt_ddb.h"
67#include "opt_mac.h"
68
69#ifdef DDB
70#include <ddb/ddb.h>
71#endif
72
73#define	ROOTNAME		"root_device"
74#define	VFS_MOUNTARG_SIZE_MAX	(1024 * 64)
75
76static int	vfs_domount(struct thread *td, const char *fstype,
77		    char *fspath, int fsflags, void *fsdata);
78static int	vfs_mount_alloc(struct vnode *dvp, struct vfsconf *vfsp,
79		    const char *fspath, struct thread *td, struct mount **mpp);
80static int	vfs_mountroot_ask(void);
81static int	vfs_mountroot_try(const char *mountfrom);
82static int	vfs_donmount(struct thread *td, int fsflags,
83		    struct uio *fsoptions);
84static void	free_mntarg(struct mntarg *ma);
85static void	vfs_mount_destroy(struct mount *, struct thread *);
86
87static int	usermount = 0;
88SYSCTL_INT(_vfs, OID_AUTO, usermount, CTLFLAG_RW, &usermount, 0,
89    "Unprivileged users may mount and unmount file systems");
90
91MALLOC_DEFINE(M_MOUNT, "mount", "vfs mount structure");
92
93/* List of mounted filesystems. */
94struct mntlist mountlist = TAILQ_HEAD_INITIALIZER(mountlist);
95
96/* For any iteration/modification of mountlist */
97struct mtx mountlist_mtx;
98MTX_SYSINIT(mountlist, &mountlist_mtx, "mountlist", MTX_DEF);
99
100TAILQ_HEAD(vfsoptlist, vfsopt);
101struct vfsopt {
102	TAILQ_ENTRY(vfsopt) link;
103	char	*name;
104	void	*value;
105	int	len;
106};
107
108/*
109 * The vnode of the system's root (/ in the filesystem, without chroot
110 * active.)
111 */
112struct vnode	*rootvnode;
113
114/*
115 * The root filesystem is detailed in the kernel environment variable
116 * vfs.root.mountfrom, which is expected to be in the general format
117 *
118 * <vfsname>:[<path>]
119 * vfsname   := the name of a VFS known to the kernel and capable
120 *              of being mounted as root
121 * path      := disk device name or other data used by the filesystem
122 *              to locate its physical store
123 */
124
125/*
126 * Global opts, taken by all filesystems
127 */
128static const char *global_opts[] = {
129	"fstype",
130	"fspath",
131	"ro",
132	"suid",
133	"exec",
134	NULL
135};
136
137/*
138 * The root specifiers we will try if RB_CDROM is specified.
139 */
140static char *cdrom_rootdevnames[] = {
141	"cd9660:cd0",
142	"cd9660:acd0",
143	NULL
144};
145
146/* legacy find-root code */
147char		*rootdevnames[2] = {NULL, NULL};
148#ifndef ROOTDEVNAME
149#  define ROOTDEVNAME NULL
150#endif
151static const char	*ctrootdevname = ROOTDEVNAME;
152
153/*
154 * ---------------------------------------------------------------------
155 * Functions for building and sanitizing the mount options
156 */
157
158/* Remove one mount option. */
159static void
160vfs_freeopt(struct vfsoptlist *opts, struct vfsopt *opt)
161{
162
163	TAILQ_REMOVE(opts, opt, link);
164	free(opt->name, M_MOUNT);
165	if (opt->value != NULL)
166		free(opt->value, M_MOUNT);
167#ifdef INVARIANTS
168	else if (opt->len != 0)
169		panic("%s: mount option with NULL value but length != 0",
170		    __func__);
171#endif
172	free(opt, M_MOUNT);
173}
174
175/* Release all resources related to the mount options. */
176static void
177vfs_freeopts(struct vfsoptlist *opts)
178{
179	struct vfsopt *opt;
180
181	while (!TAILQ_EMPTY(opts)) {
182		opt = TAILQ_FIRST(opts);
183		vfs_freeopt(opts, opt);
184	}
185	free(opts, M_MOUNT);
186}
187
188/*
189 * Check if options are equal (with or without the "no" prefix).
190 */
191static int
192vfs_equalopts(const char *opt1, const char *opt2)
193{
194
195	/* "opt" vs. "opt" or "noopt" vs. "noopt" */
196	if (strcmp(opt1, opt2) == 0)
197		return (1);
198	/* "noopt" vs. "opt" */
199	if (strncmp(opt1, "no", 2) == 0 && strcmp(opt1 + 2, opt2) == 0)
200		return (1);
201	/* "opt" vs. "noopt" */
202	if (strncmp(opt2, "no", 2) == 0 && strcmp(opt1, opt2 + 2) == 0)
203		return (1);
204	return (0);
205}
206
207/*
208 * If a mount option is specified several times,
209 * (with or without the "no" prefix) only keep
210 * the last occurence of it.
211 */
212static void
213vfs_sanitizeopts(struct vfsoptlist *opts)
214{
215	struct vfsopt *opt, *opt2, *tmp;
216
217	TAILQ_FOREACH_REVERSE(opt, opts, vfsoptlist, link) {
218		opt2 = TAILQ_PREV(opt, vfsoptlist, link);
219		while (opt2 != NULL) {
220			if (vfs_equalopts(opt->name, opt2->name)) {
221				tmp = TAILQ_PREV(opt2, vfsoptlist, link);
222				vfs_freeopt(opts, opt2);
223				opt2 = tmp;
224			} else {
225				opt2 = TAILQ_PREV(opt2, vfsoptlist, link);
226			}
227		}
228	}
229}
230
231/*
232 * Build a linked list of mount options from a struct uio.
233 */
234static int
235vfs_buildopts(struct uio *auio, struct vfsoptlist **options)
236{
237	struct vfsoptlist *opts;
238	struct vfsopt *opt;
239	size_t memused;
240	unsigned int i, iovcnt;
241	int error, namelen, optlen;
242
243	opts = malloc(sizeof(struct vfsoptlist), M_MOUNT, M_WAITOK);
244	TAILQ_INIT(opts);
245	memused = 0;
246	iovcnt = auio->uio_iovcnt;
247	for (i = 0; i < iovcnt; i += 2) {
248		opt = malloc(sizeof(struct vfsopt), M_MOUNT, M_WAITOK);
249		namelen = auio->uio_iov[i].iov_len;
250		optlen = auio->uio_iov[i + 1].iov_len;
251		opt->name = malloc(namelen, M_MOUNT, M_WAITOK);
252		opt->value = NULL;
253		opt->len = 0;
254
255		/*
256		 * Do this early, so jumps to "bad" will free the current
257		 * option.
258		 */
259		TAILQ_INSERT_TAIL(opts, opt, link);
260		memused += sizeof(struct vfsopt) + optlen + namelen;
261
262		/*
263		 * Avoid consuming too much memory, and attempts to overflow
264		 * memused.
265		 */
266		if (memused > VFS_MOUNTARG_SIZE_MAX ||
267		    optlen > VFS_MOUNTARG_SIZE_MAX ||
268		    namelen > VFS_MOUNTARG_SIZE_MAX) {
269			error = EINVAL;
270			goto bad;
271		}
272
273		if (auio->uio_segflg == UIO_SYSSPACE) {
274			bcopy(auio->uio_iov[i].iov_base, opt->name, namelen);
275		} else {
276			error = copyin(auio->uio_iov[i].iov_base, opt->name,
277			    namelen);
278			if (error)
279				goto bad;
280		}
281		/* Ensure names are null-terminated strings. */
282		if (opt->name[namelen - 1] != '\0') {
283			error = EINVAL;
284			goto bad;
285		}
286		if (optlen != 0) {
287			opt->len = optlen;
288			opt->value = malloc(optlen, M_MOUNT, M_WAITOK);
289			if (auio->uio_segflg == UIO_SYSSPACE) {
290				bcopy(auio->uio_iov[i + 1].iov_base, opt->value,
291				    optlen);
292			} else {
293				error = copyin(auio->uio_iov[i + 1].iov_base,
294				    opt->value, optlen);
295				if (error)
296					goto bad;
297			}
298		}
299	}
300	vfs_sanitizeopts(opts);
301	*options = opts;
302	return (0);
303bad:
304	vfs_freeopts(opts);
305	return (error);
306}
307
308/*
309 * Merge the old mount options with the new ones passed
310 * in the MNT_UPDATE case.
311 */
312static void
313vfs_mergeopts(struct vfsoptlist *toopts, struct vfsoptlist *opts)
314{
315	struct vfsopt *opt, *opt2, *new;
316
317	TAILQ_FOREACH(opt, opts, link) {
318		/*
319		 * Check that this option hasn't been redefined
320		 * nor cancelled with a "no" mount option.
321		 */
322		opt2 = TAILQ_FIRST(toopts);
323		while (opt2 != NULL) {
324			if (strcmp(opt2->name, opt->name) == 0)
325				goto next;
326			if (strncmp(opt2->name, "no", 2) == 0 &&
327			    strcmp(opt2->name + 2, opt->name) == 0) {
328				vfs_freeopt(toopts, opt2);
329				goto next;
330			}
331			opt2 = TAILQ_NEXT(opt2, link);
332		}
333		/* We want this option, duplicate it. */
334		new = malloc(sizeof(struct vfsopt), M_MOUNT, M_WAITOK);
335		new->name = malloc(strlen(opt->name) + 1, M_MOUNT, M_WAITOK);
336		strcpy(new->name, opt->name);
337		if (opt->len != 0) {
338			new->value = malloc(opt->len, M_MOUNT, M_WAITOK);
339			bcopy(opt->value, new->value, opt->len);
340		} else {
341			new->value = NULL;
342		}
343		new->len = opt->len;
344		TAILQ_INSERT_TAIL(toopts, new, link);
345next:
346		continue;
347	}
348}
349
350/*
351 * ---------------------------------------------------------------------
352 * Mount a filesystem
353 */
354int
355nmount(td, uap)
356	struct thread *td;
357	struct nmount_args /* {
358		struct iovec *iovp;
359		unsigned int iovcnt;
360		int flags;
361	} */ *uap;
362{
363	struct uio *auio;
364	struct iovec *iov;
365	unsigned int i;
366	int error;
367	u_int iovcnt;
368
369	/* Kick out MNT_ROOTFS early as it is legal internally */
370	if (uap->flags & MNT_ROOTFS)
371		return (EINVAL);
372
373	iovcnt = uap->iovcnt;
374	/*
375	 * Check that we have an even number of iovec's
376	 * and that we have at least two options.
377	 */
378	if ((iovcnt & 1) || (iovcnt < 4))
379		return (EINVAL);
380
381	error = copyinuio(uap->iovp, iovcnt, &auio);
382	if (error)
383		return (error);
384	iov = auio->uio_iov;
385	for (i = 0; i < iovcnt; i++) {
386		if (iov->iov_len > MMAXOPTIONLEN) {
387			free(auio, M_IOV);
388			return (EINVAL);
389		}
390		iov++;
391	}
392	error = vfs_donmount(td, uap->flags, auio);
393	free(auio, M_IOV);
394	return (error);
395}
396
397/*
398 * ---------------------------------------------------------------------
399 * Various utility functions
400 */
401
402/*
403 * Allocate and initialize the mount point struct.
404 */
405static int
406vfs_mount_alloc(struct vnode *vp, struct vfsconf *vfsp,
407    const char *fspath, struct thread *td, struct mount **mpp)
408{
409	struct mount *mp;
410
411	mp = malloc(sizeof(struct mount), M_MOUNT, M_WAITOK | M_ZERO);
412	TAILQ_INIT(&mp->mnt_nvnodelist);
413	mp->mnt_nvnodelistsize = 0;
414	mtx_init(&mp->mnt_mtx, "struct mount mtx", NULL, MTX_DEF);
415	lockinit(&mp->mnt_lock, PVFS, "vfslock", 0, 0);
416	vfs_busy(mp, LK_NOWAIT, 0, td);
417	mp->mnt_op = vfsp->vfc_vfsops;
418	mp->mnt_vfc = vfsp;
419	vfsp->vfc_refcount++;
420	mp->mnt_stat.f_type = vfsp->vfc_typenum;
421	mp->mnt_flag |= vfsp->vfc_flags & MNT_VISFLAGMASK;
422	strlcpy(mp->mnt_stat.f_fstypename, vfsp->vfc_name, MFSNAMELEN);
423	mp->mnt_vnodecovered = vp;
424	mp->mnt_cred = crdup(td->td_ucred);
425	mp->mnt_stat.f_owner = td->td_ucred->cr_uid;
426	strlcpy(mp->mnt_stat.f_mntonname, fspath, MNAMELEN);
427	mp->mnt_iosize_max = DFLTPHYS;
428#ifdef MAC
429	mac_init_mount(mp);
430	mac_create_mount(td->td_ucred, mp);
431#endif
432	arc4rand(&mp->mnt_hashseed, sizeof mp->mnt_hashseed, 0);
433	*mpp = mp;
434	return (0);
435}
436
437/*
438 * Destroy the mount struct previously allocated by vfs_mount_alloc().
439 */
440static void
441vfs_mount_destroy(struct mount *mp, struct thread *td)
442{
443
444	mp->mnt_vfc->vfc_refcount--;
445	if (!TAILQ_EMPTY(&mp->mnt_nvnodelist))
446		panic("unmount: dangling vnode");
447	vfs_unbusy(mp,td);
448	lockdestroy(&mp->mnt_lock);
449	MNT_ILOCK(mp);
450	if (mp->mnt_kern_flag & MNTK_MWAIT)
451		wakeup(mp);
452	MNT_IUNLOCK(mp);
453	mtx_destroy(&mp->mnt_mtx);
454#ifdef MAC
455	mac_destroy_mount(mp);
456#endif
457	if (mp->mnt_opt != NULL)
458		vfs_freeopts(mp->mnt_opt);
459	crfree(mp->mnt_cred);
460	free(mp, M_MOUNT);
461}
462
463static int
464vfs_donmount(struct thread *td, int fsflags, struct uio *fsoptions)
465{
466	struct vfsoptlist *optlist;
467	char *fstype, *fspath;
468	int error, fstypelen, fspathlen;
469
470	error = vfs_buildopts(fsoptions, &optlist);
471	if (error)
472		return (error);
473
474	/*
475	 * We need these two options before the others,
476	 * and they are mandatory for any filesystem.
477	 * Ensure they are NUL terminated as well.
478	 */
479	fstypelen = 0;
480	error = vfs_getopt(optlist, "fstype", (void **)&fstype, &fstypelen);
481	if (error || fstype[fstypelen - 1] != '\0') {
482		error = EINVAL;
483		goto bail;
484	}
485	fspathlen = 0;
486	error = vfs_getopt(optlist, "fspath", (void **)&fspath, &fspathlen);
487	if (error || fspath[fspathlen - 1] != '\0') {
488		error = EINVAL;
489		goto bail;
490	}
491
492	/*
493	 * Be ultra-paranoid about making sure the type and fspath
494	 * variables will fit in our mp buffers, including the
495	 * terminating NUL.
496	 */
497	if (fstypelen >= MFSNAMELEN - 1 || fspathlen >= MNAMELEN - 1) {
498		error = ENAMETOOLONG;
499		goto bail;
500	}
501
502	mtx_lock(&Giant);
503	error = vfs_domount(td, fstype, fspath, fsflags, optlist);
504	mtx_unlock(&Giant);
505bail:
506	if (error)
507		vfs_freeopts(optlist);
508	return (error);
509}
510
511/*
512 * ---------------------------------------------------------------------
513 * Old mount API.
514 */
515#ifndef _SYS_SYSPROTO_H_
516struct mount_args {
517	char	*type;
518	char	*path;
519	int	flags;
520	caddr_t	data;
521};
522#endif
523/* ARGSUSED */
524int
525mount(td, uap)
526	struct thread *td;
527	struct mount_args /* {
528		char *type;
529		char *path;
530		int flags;
531		caddr_t data;
532	} */ *uap;
533{
534	char *fstype;
535	struct vfsconf *vfsp = NULL;
536	struct mntarg *ma = NULL;
537	int error;
538
539	/* Kick out MNT_ROOTFS early as it is legal internally */
540	uap->flags &= ~MNT_ROOTFS;
541
542	if (uap->data == NULL)
543		return (EINVAL);
544
545	fstype = malloc(MFSNAMELEN, M_TEMP, M_WAITOK);
546	error = copyinstr(uap->type, fstype, MFSNAMELEN, NULL);
547	if (!error) {
548		mtx_lock(&Giant);	/* XXX ? */
549		vfsp = vfs_byname_kld(fstype, td, &error);
550		mtx_unlock(&Giant);
551	}
552	free(fstype, M_TEMP);
553	if (error)
554		return (error);
555	if (vfsp == NULL)
556		return (ENOENT);
557	if (vfsp->vfc_vfsops->vfs_cmount == NULL)
558		return (EOPNOTSUPP);
559
560	ma = mount_argsu(ma, "fstype", uap->type, MNAMELEN);
561	ma = mount_argsu(ma, "fspath", uap->path, MNAMELEN);
562	ma = mount_argb(ma, uap->flags & MNT_RDONLY, "noro");
563	ma = mount_argb(ma, !(uap->flags & MNT_NOSUID), "nosuid");
564	ma = mount_argb(ma, !(uap->flags & MNT_NOEXEC), "noexec");
565
566	error = vfsp->vfc_vfsops->vfs_cmount(ma, uap->data, uap->flags, td);
567	return (error);
568}
569
570
571/*
572 * vfs_domount(): actually attempt a filesystem mount.
573 */
574static int
575vfs_domount(
576	struct thread *td,	/* Flags common to all filesystems. */
577	const char *fstype,	/* Filesystem type. */
578	char *fspath,		/* Mount path. */
579	int fsflags,		/* Flags common to all filesystems. */
580	void *fsdata		/* Options local to the filesystem. */
581	)
582{
583	struct vnode *vp;
584	struct mount *mp;
585	struct vfsconf *vfsp;
586	int error, flag = 0, kern_flag = 0;
587	struct vattr va;
588	struct nameidata nd;
589
590	mtx_assert(&Giant, MA_OWNED);
591
592	/*
593	 * Be ultra-paranoid about making sure the type and fspath
594	 * variables will fit in our mp buffers, including the
595	 * terminating NUL.
596	 */
597	if (strlen(fstype) >= MFSNAMELEN || strlen(fspath) >= MNAMELEN)
598		return (ENAMETOOLONG);
599
600	if (jailed(td->td_ucred))
601		return (EPERM);
602	if (usermount == 0) {
603		if ((error = suser(td)) != 0)
604			return (error);
605	}
606
607	/*
608	 * Do not allow NFS export or MNT_SUIDDIR by unprivileged users.
609	 */
610	if (fsflags & (MNT_EXPORTED | MNT_SUIDDIR)) {
611		if ((error = suser(td)) != 0)
612			return (error);
613	}
614	/*
615	 * Silently enforce MNT_NOSUID and MNT_USER for
616	 * unprivileged users.
617	 */
618	if (suser(td) != 0)
619		fsflags |= MNT_NOSUID | MNT_USER;
620	/*
621	 * Get vnode to be covered
622	 */
623	NDINIT(&nd, LOOKUP, FOLLOW | LOCKLEAF, UIO_SYSSPACE, fspath, td);
624	if ((error = namei(&nd)) != 0)
625		return (error);
626	NDFREE(&nd, NDF_ONLY_PNBUF);
627	vp = nd.ni_vp;
628	if (fsflags & MNT_UPDATE) {
629		if ((vp->v_vflag & VV_ROOT) == 0) {
630			vput(vp);
631			return (EINVAL);
632		}
633		mp = vp->v_mount;
634		flag = mp->mnt_flag;
635		kern_flag = mp->mnt_kern_flag;
636		/*
637		 * We only allow the filesystem to be reloaded if it
638		 * is currently mounted read-only.
639		 */
640		if ((fsflags & MNT_RELOAD) &&
641		    ((mp->mnt_flag & MNT_RDONLY) == 0)) {
642			vput(vp);
643			return (EOPNOTSUPP);	/* Needs translation */
644		}
645		/*
646		 * Only privileged root, or (if MNT_USER is set) the user that
647		 * did the original mount is permitted to update it.
648		 */
649		error = vfs_suser(mp, td);
650		if (error) {
651			vput(vp);
652			return (error);
653		}
654		if (vfs_busy(mp, LK_NOWAIT, 0, td)) {
655			vput(vp);
656			return (EBUSY);
657		}
658		VI_LOCK(vp);
659		if ((vp->v_iflag & VI_MOUNT) != 0 ||
660		    vp->v_mountedhere != NULL) {
661			VI_UNLOCK(vp);
662			vfs_unbusy(mp, td);
663			vput(vp);
664			return (EBUSY);
665		}
666		vp->v_iflag |= VI_MOUNT;
667		VI_UNLOCK(vp);
668		mp->mnt_flag |= fsflags &
669		    (MNT_RELOAD | MNT_FORCE | MNT_UPDATE | MNT_SNAPSHOT | MNT_ROOTFS);
670		VOP_UNLOCK(vp, 0, td);
671		mp->mnt_optnew = fsdata;
672		vfs_mergeopts(mp->mnt_optnew, mp->mnt_opt);
673	} else {
674		/*
675		 * If the user is not root, ensure that they own the directory
676		 * onto which we are attempting to mount.
677		 */
678		error = VOP_GETATTR(vp, &va, td->td_ucred, td);
679		if (error) {
680			vput(vp);
681			return (error);
682		}
683		if (va.va_uid != td->td_ucred->cr_uid) {
684			if ((error = suser(td)) != 0) {
685				vput(vp);
686				return (error);
687			}
688		}
689		error = vinvalbuf(vp, V_SAVE, td, 0, 0);
690		if (error != 0) {
691			vput(vp);
692			return (error);
693		}
694		if (vp->v_type != VDIR) {
695			vput(vp);
696			return (ENOTDIR);
697		}
698		vfsp = vfs_byname_kld(fstype, td, &error);
699		if (vfsp == NULL) {
700			vput(vp);
701			return (error);
702		}
703		VI_LOCK(vp);
704		if ((vp->v_iflag & VI_MOUNT) != 0 ||
705		    vp->v_mountedhere != NULL) {
706			VI_UNLOCK(vp);
707			vput(vp);
708			return (EBUSY);
709		}
710		vp->v_iflag |= VI_MOUNT;
711		VI_UNLOCK(vp);
712
713		/*
714		 * Allocate and initialize the filesystem.
715		 */
716		error = vfs_mount_alloc(vp, vfsp, fspath, td, &mp);
717		if (error) {
718			vput(vp);
719			return (error);
720		}
721		VOP_UNLOCK(vp, 0, td);
722
723		/* XXXMAC: pass to vfs_mount_alloc? */
724		mp->mnt_optnew = fsdata;
725	}
726
727	/*
728	 * Set the mount level flags.
729	 */
730	if (fsflags & MNT_RDONLY)
731		mp->mnt_flag |= MNT_RDONLY;
732	mp->mnt_flag &=~ MNT_UPDATEMASK;
733	mp->mnt_flag |= fsflags & (MNT_UPDATEMASK | MNT_FORCE | MNT_ROOTFS);
734	/*
735	 * Mount the filesystem.
736	 * XXX The final recipients of VFS_MOUNT just overwrite the ndp they
737	 * get.  No freeing of cn_pnbuf.
738	 */
739        error = VFS_MOUNT(mp, td);
740	if (!error) {
741		if (mp->mnt_opt != NULL)
742			vfs_freeopts(mp->mnt_opt);
743		mp->mnt_opt = mp->mnt_optnew;
744		VFS_STATFS(mp, &mp->mnt_stat, td);
745	}
746	/*
747	 * Prevent external consumers of mount options from reading
748	 * mnt_optnew.
749	*/
750	mp->mnt_optnew = NULL;
751	if (mp->mnt_flag & MNT_UPDATE) {
752		mp->mnt_flag &=
753		    ~(MNT_UPDATE | MNT_RELOAD | MNT_FORCE | MNT_SNAPSHOT);
754		if (error) {
755			mp->mnt_flag = flag;
756			mp->mnt_kern_flag = kern_flag;
757		}
758		if ((mp->mnt_flag & MNT_RDONLY) == 0) {
759			if (mp->mnt_syncer == NULL)
760				error = vfs_allocate_syncvnode(mp);
761		} else {
762			if (mp->mnt_syncer != NULL)
763				vrele(mp->mnt_syncer);
764			mp->mnt_syncer = NULL;
765		}
766		vfs_unbusy(mp, td);
767		VI_LOCK(vp);
768		vp->v_iflag &= ~VI_MOUNT;
769		VI_UNLOCK(vp);
770		vrele(vp);
771		return (error);
772	}
773	vn_lock(vp, LK_EXCLUSIVE | LK_RETRY, td);
774	/*
775	 * Put the new filesystem on the mount list after root.
776	 */
777	cache_purge(vp);
778	if (!error) {
779		struct vnode *newdp;
780
781		VI_LOCK(vp);
782		vp->v_iflag &= ~VI_MOUNT;
783		VI_UNLOCK(vp);
784		vp->v_mountedhere = mp;
785		mtx_lock(&mountlist_mtx);
786		TAILQ_INSERT_TAIL(&mountlist, mp, mnt_list);
787		mtx_unlock(&mountlist_mtx);
788		vfs_event_signal(NULL, VQ_MOUNT, 0);
789		if (VFS_ROOT(mp, LK_EXCLUSIVE, &newdp, td))
790			panic("mount: lost mount");
791		mountcheckdirs(vp, newdp);
792		vput(newdp);
793		VOP_UNLOCK(vp, 0, td);
794		if ((mp->mnt_flag & MNT_RDONLY) == 0)
795			error = vfs_allocate_syncvnode(mp);
796		vfs_unbusy(mp, td);
797		if (error)
798			vrele(vp);
799	} else {
800		VI_LOCK(vp);
801		vp->v_iflag &= ~VI_MOUNT;
802		VI_UNLOCK(vp);
803		vfs_mount_destroy(mp, td);
804		vput(vp);
805	}
806	return (error);
807}
808
809/*
810 * ---------------------------------------------------------------------
811 * Unmount a filesystem.
812 *
813 * Note: unmount takes a path to the vnode mounted on as argument,
814 * not special file (as before).
815 */
816#ifndef _SYS_SYSPROTO_H_
817struct unmount_args {
818	char	*path;
819	int	flags;
820};
821#endif
822/* ARGSUSED */
823int
824unmount(td, uap)
825	struct thread *td;
826	register struct unmount_args /* {
827		char *path;
828		int flags;
829	} */ *uap;
830{
831	struct mount *mp;
832	char *pathbuf;
833	int error, id0, id1;
834
835	if (jailed(td->td_ucred))
836		return (EPERM);
837	if (usermount == 0) {
838		if ((error = suser(td)) != 0)
839			return (error);
840	}
841
842	pathbuf = malloc(MNAMELEN, M_TEMP, M_WAITOK);
843	error = copyinstr(uap->path, pathbuf, MNAMELEN, NULL);
844	if (error) {
845		free(pathbuf, M_TEMP);
846		return (error);
847	}
848	if (uap->flags & MNT_BYFSID) {
849		/* Decode the filesystem ID. */
850		if (sscanf(pathbuf, "FSID:%d:%d", &id0, &id1) != 2) {
851			free(pathbuf, M_TEMP);
852			return (EINVAL);
853		}
854
855		mtx_lock(&mountlist_mtx);
856		TAILQ_FOREACH_REVERSE(mp, &mountlist, mntlist, mnt_list) {
857			if (mp->mnt_stat.f_fsid.val[0] == id0 &&
858			    mp->mnt_stat.f_fsid.val[1] == id1)
859				break;
860		}
861		mtx_unlock(&mountlist_mtx);
862	} else {
863		mtx_lock(&mountlist_mtx);
864		TAILQ_FOREACH_REVERSE(mp, &mountlist, mntlist, mnt_list) {
865			if (strcmp(mp->mnt_stat.f_mntonname, pathbuf) == 0)
866				break;
867		}
868		mtx_unlock(&mountlist_mtx);
869	}
870	free(pathbuf, M_TEMP);
871	if (mp == NULL) {
872		/*
873		 * Previously we returned ENOENT for a nonexistent path and
874		 * EINVAL for a non-mountpoint.  We cannot tell these apart
875		 * now, so in the !MNT_BYFSID case return the more likely
876		 * EINVAL for compatibility.
877		 */
878		return ((uap->flags & MNT_BYFSID) ? ENOENT : EINVAL);
879	}
880
881	/*
882	 * Only privileged root, or (if MNT_USER is set) the user that did the
883	 * original mount is permitted to unmount this filesystem.
884	 */
885	error = vfs_suser(mp, td);
886	if (error)
887		return (error);
888
889	/*
890	 * Don't allow unmounting the root filesystem.
891	 */
892	if (mp->mnt_flag & MNT_ROOTFS)
893		return (EINVAL);
894	mtx_lock(&Giant);
895	error = dounmount(mp, uap->flags, td);
896	mtx_unlock(&Giant);
897	return (error);
898}
899
900/*
901 * Do the actual filesystem unmount.
902 */
903int
904dounmount(mp, flags, td)
905	struct mount *mp;
906	int flags;
907	struct thread *td;
908{
909	struct vnode *coveredvp, *fsrootvp;
910	int error;
911	int async_flag;
912
913	mtx_assert(&Giant, MA_OWNED);
914
915	MNT_ILOCK(mp);
916	if (mp->mnt_kern_flag & MNTK_UNMOUNT) {
917		MNT_IUNLOCK(mp);
918		return (EBUSY);
919	}
920	mp->mnt_kern_flag |= MNTK_UNMOUNT;
921	/* Allow filesystems to detect that a forced unmount is in progress. */
922	if (flags & MNT_FORCE)
923		mp->mnt_kern_flag |= MNTK_UNMOUNTF;
924	error = lockmgr(&mp->mnt_lock, LK_DRAIN | LK_INTERLOCK |
925	    ((flags & MNT_FORCE) ? 0 : LK_NOWAIT), MNT_MTX(mp), td);
926	if (error) {
927		MNT_ILOCK(mp);
928		mp->mnt_kern_flag &= ~(MNTK_UNMOUNT | MNTK_UNMOUNTF);
929		if (mp->mnt_kern_flag & MNTK_MWAIT)
930			wakeup(mp);
931		MNT_IUNLOCK(mp);
932		return (error);
933	}
934	vn_start_write(NULL, &mp, V_WAIT);
935
936	if (mp->mnt_flag & MNT_EXPUBLIC)
937		vfs_setpublicfs(NULL, NULL, NULL);
938
939	vfs_msync(mp, MNT_WAIT);
940	async_flag = mp->mnt_flag & MNT_ASYNC;
941	mp->mnt_flag &= ~MNT_ASYNC;
942	cache_purgevfs(mp);	/* remove cache entries for this file sys */
943	if (mp->mnt_syncer != NULL)
944		vrele(mp->mnt_syncer);
945	/*
946	 * For forced unmounts, move process cdir/rdir refs on the fs root
947	 * vnode to the covered vnode.  For non-forced unmounts we want
948	 * such references to cause an EBUSY error.
949	 */
950	if ((flags & MNT_FORCE) &&
951	    VFS_ROOT(mp, LK_EXCLUSIVE, &fsrootvp, td) == 0) {
952		if (mp->mnt_vnodecovered != NULL)
953			mountcheckdirs(fsrootvp, mp->mnt_vnodecovered);
954		if (fsrootvp == rootvnode) {
955			vrele(rootvnode);
956			rootvnode = NULL;
957		}
958		vput(fsrootvp);
959	}
960	if (((mp->mnt_flag & MNT_RDONLY) ||
961	     (error = VFS_SYNC(mp, MNT_WAIT, td)) == 0) ||
962	    (flags & MNT_FORCE)) {
963		error = VFS_UNMOUNT(mp, flags, td);
964	}
965	vn_finished_write(mp);
966	if (error) {
967		/* Undo cdir/rdir and rootvnode changes made above. */
968		if ((flags & MNT_FORCE) &&
969		    VFS_ROOT(mp, LK_EXCLUSIVE, &fsrootvp, td) == 0) {
970			if (mp->mnt_vnodecovered != NULL)
971				mountcheckdirs(mp->mnt_vnodecovered, fsrootvp);
972			if (rootvnode == NULL) {
973				rootvnode = fsrootvp;
974				vref(rootvnode);
975			}
976			vput(fsrootvp);
977		}
978		if ((mp->mnt_flag & MNT_RDONLY) == 0 && mp->mnt_syncer == NULL)
979			(void) vfs_allocate_syncvnode(mp);
980		MNT_ILOCK(mp);
981		mp->mnt_kern_flag &= ~(MNTK_UNMOUNT | MNTK_UNMOUNTF);
982		mp->mnt_flag |= async_flag;
983		lockmgr(&mp->mnt_lock, LK_RELEASE, NULL, td);
984		if (mp->mnt_kern_flag & MNTK_MWAIT)
985			wakeup(mp);
986		MNT_IUNLOCK(mp);
987		return (error);
988	}
989	mtx_lock(&mountlist_mtx);
990	TAILQ_REMOVE(&mountlist, mp, mnt_list);
991	if ((coveredvp = mp->mnt_vnodecovered) != NULL)
992		coveredvp->v_mountedhere = NULL;
993	mtx_unlock(&mountlist_mtx);
994	vfs_event_signal(NULL, VQ_UNMOUNT, 0);
995	vfs_mount_destroy(mp, td);
996	if (coveredvp != NULL)
997		vrele(coveredvp);
998	return (0);
999}
1000
1001/*
1002 * ---------------------------------------------------------------------
1003 * Mounting of root filesystem
1004 *
1005 */
1006
1007struct root_hold_token {
1008	const char 			*who;
1009	LIST_ENTRY(root_hold_token)	list;
1010};
1011
1012static LIST_HEAD(, root_hold_token)	root_holds =
1013    LIST_HEAD_INITIALIZER(&root_holds);
1014
1015struct root_hold_token *
1016root_mount_hold(const char *identifier)
1017{
1018	struct root_hold_token *h;
1019
1020	h = malloc(sizeof *h, M_DEVBUF, M_ZERO | M_WAITOK);
1021	h->who = identifier;
1022	mtx_lock(&mountlist_mtx);
1023	LIST_INSERT_HEAD(&root_holds, h, list);
1024	mtx_unlock(&mountlist_mtx);
1025	return (h);
1026}
1027
1028void
1029root_mount_rel(struct root_hold_token *h)
1030{
1031
1032	mtx_lock(&mountlist_mtx);
1033	LIST_REMOVE(h, list);
1034	wakeup(&root_holds);
1035	mtx_unlock(&mountlist_mtx);
1036	free(h, M_DEVBUF);
1037}
1038
1039static void
1040root_mount_wait(void)
1041{
1042	struct root_hold_token *h;
1043
1044	for (;;) {
1045		DROP_GIANT();
1046		g_waitidle();
1047		PICKUP_GIANT();
1048		mtx_lock(&mountlist_mtx);
1049		if (LIST_EMPTY(&root_holds)) {
1050			mtx_unlock(&mountlist_mtx);
1051			break;
1052		}
1053		printf("Root mount waiting for:");
1054		LIST_FOREACH(h, &root_holds, list)
1055			printf(" %s", h->who);
1056		printf("\n");
1057		msleep(&root_holds, &mountlist_mtx, PZERO | PDROP, "roothold",
1058		    hz);
1059	}
1060}
1061
1062static void
1063set_rootvnode(struct thread *td)
1064{
1065	struct proc *p;
1066
1067	if (VFS_ROOT(TAILQ_FIRST(&mountlist), LK_EXCLUSIVE, &rootvnode, td))
1068		panic("Cannot find root vnode");
1069
1070	p = td->td_proc;
1071	FILEDESC_LOCK(p->p_fd);
1072
1073	if (p->p_fd->fd_cdir != NULL)
1074		vrele(p->p_fd->fd_cdir);
1075	p->p_fd->fd_cdir = rootvnode;
1076	VREF(rootvnode);
1077
1078	if (p->p_fd->fd_rdir != NULL)
1079		vrele(p->p_fd->fd_rdir);
1080	p->p_fd->fd_rdir = rootvnode;
1081	VREF(rootvnode);
1082
1083	FILEDESC_UNLOCK(p->p_fd);
1084
1085	VOP_UNLOCK(rootvnode, 0, td);
1086}
1087
1088/*
1089 * Mount /devfs as our root filesystem, but do not put it on the mountlist
1090 * yet.  Create a /dev -> / symlink so that absolute pathnames will lookup.
1091 */
1092
1093static struct mount *
1094devfs_first(void)
1095{
1096	struct thread *td = curthread;
1097	struct vfsconf *vfsp;
1098	struct mount *mp = NULL;
1099	int error;
1100
1101	vfsp = vfs_byname("devfs");
1102	KASSERT(vfsp != NULL, ("Could not find devfs by name"));
1103	if (vfsp == NULL)
1104		return(NULL);
1105
1106	error = vfs_mount_alloc(NULLVP, vfsp, "/dev", td, &mp);
1107	KASSERT(error == 0, ("vfs_mount_alloc failed %d", error));
1108	if (error)
1109		return (NULL);
1110
1111	error = VFS_MOUNT(mp, curthread);
1112	KASSERT(error == 0, ("VFS_MOUNT(devfs) failed %d", error));
1113	if (error)
1114		return (NULL);
1115
1116	mtx_lock(&mountlist_mtx);
1117	TAILQ_INSERT_HEAD(&mountlist, mp, mnt_list);
1118	mtx_unlock(&mountlist_mtx);
1119
1120	set_rootvnode(td);
1121
1122	error = kern_symlink(td, "/", "dev", UIO_SYSSPACE);
1123	if (error)
1124		printf("kern_symlink /dev -> / returns %d\n", error);
1125
1126	return (mp);
1127}
1128
1129/*
1130 * Surgically move our devfs to be mounted on /dev.
1131 */
1132
1133static void
1134devfs_fixup(struct thread *td)
1135{
1136	struct nameidata nd;
1137	int error;
1138	struct vnode *vp, *dvp;
1139	struct mount *mp;
1140
1141	/* Remove our devfs mount from the mountlist and purge the cache */
1142	mtx_lock(&mountlist_mtx);
1143	mp = TAILQ_FIRST(&mountlist);
1144	TAILQ_REMOVE(&mountlist, mp, mnt_list);
1145	mtx_unlock(&mountlist_mtx);
1146	cache_purgevfs(mp);
1147
1148	VFS_ROOT(mp, LK_EXCLUSIVE, &dvp, td);
1149	VI_LOCK(dvp);
1150	dvp->v_iflag &= ~VI_MOUNT;
1151	dvp->v_mountedhere = NULL;
1152	VI_UNLOCK(dvp);
1153
1154	/* Set up the real rootvnode, and purge the cache */
1155	TAILQ_FIRST(&mountlist)->mnt_vnodecovered = NULL;
1156	set_rootvnode(td);
1157	cache_purgevfs(rootvnode->v_mount);
1158
1159	NDINIT(&nd, LOOKUP, FOLLOW | LOCKLEAF, UIO_SYSSPACE, "/dev", td);
1160	error = namei(&nd);
1161	if (error) {
1162		printf("Lookup of /dev for devfs, error: %d\n", error);
1163		return;
1164	}
1165	NDFREE(&nd, NDF_ONLY_PNBUF);
1166	vp = nd.ni_vp;
1167	if (vp->v_type != VDIR) {
1168		vput(vp);
1169	}
1170	error = vinvalbuf(vp, V_SAVE, td, 0, 0);
1171	if (error) {
1172		vput(vp);
1173	}
1174	cache_purge(vp);
1175	mp->mnt_vnodecovered = vp;
1176	vp->v_mountedhere = mp;
1177	mtx_lock(&mountlist_mtx);
1178	TAILQ_INSERT_TAIL(&mountlist, mp, mnt_list);
1179	mtx_unlock(&mountlist_mtx);
1180	VOP_UNLOCK(vp, 0, td);
1181	vfs_unbusy(mp, td);
1182	vput(dvp);
1183
1184	/* Unlink the no longer needed /dev/dev -> / symlink */
1185	kern_unlink(td, "/dev/dev", UIO_SYSSPACE);
1186}
1187
1188/*
1189 * Find and mount the root filesystem
1190 */
1191void
1192vfs_mountroot(void)
1193{
1194	char *cp;
1195	int error, i, asked = 0;
1196
1197	root_mount_wait();
1198
1199	devfs_first();
1200
1201	/*
1202	 * We are booted with instructions to prompt for the root filesystem.
1203	 */
1204	if (boothowto & RB_ASKNAME) {
1205		if (!vfs_mountroot_ask())
1206			return;
1207		asked = 1;
1208	}
1209
1210	/*
1211	 * The root filesystem information is compiled in, and we are
1212	 * booted with instructions to use it.
1213	 */
1214	if (ctrootdevname != NULL && (boothowto & RB_DFLTROOT)) {
1215		if (!vfs_mountroot_try(ctrootdevname))
1216			return;
1217		ctrootdevname = NULL;
1218	}
1219
1220	/*
1221	 * We've been given the generic "use CDROM as root" flag.  This is
1222	 * necessary because one media may be used in many different
1223	 * devices, so we need to search for them.
1224	 */
1225	if (boothowto & RB_CDROM) {
1226		for (i = 0; cdrom_rootdevnames[i] != NULL; i++) {
1227			if (!vfs_mountroot_try(cdrom_rootdevnames[i]))
1228				return;
1229		}
1230	}
1231
1232	/*
1233	 * Try to use the value read by the loader from /etc/fstab, or
1234	 * supplied via some other means.  This is the preferred
1235	 * mechanism.
1236	 */
1237	cp = getenv("vfs.root.mountfrom");
1238	if (cp != NULL) {
1239		error = vfs_mountroot_try(cp);
1240		freeenv(cp);
1241		if (!error)
1242			return;
1243	}
1244
1245	/*
1246	 * Try values that may have been computed by code during boot
1247	 */
1248	if (!vfs_mountroot_try(rootdevnames[0]))
1249		return;
1250	if (!vfs_mountroot_try(rootdevnames[1]))
1251		return;
1252
1253	/*
1254	 * If we (still) have a compiled-in default, try it.
1255	 */
1256	if (ctrootdevname != NULL)
1257		if (!vfs_mountroot_try(ctrootdevname))
1258			return;
1259	/*
1260	 * Everything so far has failed, prompt on the console if we haven't
1261	 * already tried that.
1262	 */
1263	if (!asked)
1264		if (!vfs_mountroot_ask())
1265			return;
1266
1267	panic("Root mount failed, startup aborted.");
1268}
1269
1270/*
1271 * Mount (mountfrom) as the root filesystem.
1272 */
1273static int
1274vfs_mountroot_try(const char *mountfrom)
1275{
1276        struct mount	*mp;
1277	char		*vfsname, *path;
1278	int		error;
1279	char		patt[32];
1280
1281	vfsname = NULL;
1282	path    = NULL;
1283	mp      = NULL;
1284	error   = EINVAL;
1285
1286	if (mountfrom == NULL)
1287		return (error);		/* don't complain */
1288	printf("Trying to mount root from %s\n", mountfrom);
1289
1290	/* parse vfs name and path */
1291	vfsname = malloc(MFSNAMELEN, M_MOUNT, M_WAITOK);
1292	path = malloc(MNAMELEN, M_MOUNT, M_WAITOK);
1293	vfsname[0] = path[0] = 0;
1294	sprintf(patt, "%%%d[a-z0-9]:%%%ds", MFSNAMELEN, MNAMELEN);
1295	if (sscanf(mountfrom, patt, vfsname, path) < 1)
1296		goto out;
1297
1298	if (path[0] == '\0')
1299		strcpy(path, ROOTNAME);
1300
1301	error = kernel_vmount(
1302	    MNT_RDONLY | MNT_ROOTFS,
1303	    "fstype", vfsname,
1304	    "fspath", "/",
1305	    "from", path,
1306	    NULL);
1307	if (error == 0) {
1308		/*
1309		 * We mount devfs prior to mounting the / FS, so the first
1310		 * entry will typically be devfs.
1311		 */
1312		mp = TAILQ_FIRST(&mountlist);
1313		KASSERT(mp != NULL, ("%s: mountlist is empty", __func__));
1314		vfs_unbusy(mp, curthread);
1315
1316		/*
1317		 * Iterate over all currently mounted file systems and use
1318		 * the time stamp found to check and/or initialize the RTC.
1319		 * Typically devfs has no time stamp and the only other FS
1320		 * is the actual / FS.
1321		 */
1322		do {
1323			if (mp->mnt_time != 0)
1324				inittodr(mp->mnt_time);
1325			mp = TAILQ_NEXT(mp, mnt_list);
1326		} while (mp != NULL);
1327
1328		devfs_fixup(curthread);
1329	}
1330out:
1331	free(path, M_MOUNT);
1332	free(vfsname, M_MOUNT);
1333	return (error);
1334}
1335
1336/*
1337 * ---------------------------------------------------------------------
1338 * Interactive root filesystem selection code.
1339 */
1340
1341static int
1342vfs_mountroot_ask(void)
1343{
1344	char name[128];
1345
1346	for(;;) {
1347		printf("\nManual root filesystem specification:\n");
1348		printf("  <fstype>:<device>  Mount <device> using filesystem <fstype>\n");
1349#if defined(__i386__) || defined(__ia64__)
1350		printf("                       eg. ufs:da0s1a\n");
1351#else
1352		printf("                       eg. ufs:/dev/da0a\n");
1353#endif
1354		printf("  ?                  List valid disk boot devices\n");
1355		printf("  <empty line>       Abort manual input\n");
1356		printf("\nmountroot> ");
1357		gets(name, sizeof(name), 1);
1358		if (name[0] == '\0')
1359			return (1);
1360		if (name[0] == '?') {
1361			printf("\nList of GEOM managed disk devices:\n  ");
1362			g_dev_print();
1363			continue;
1364		}
1365		if (!vfs_mountroot_try(name))
1366			return (0);
1367	}
1368}
1369
1370/*
1371 * ---------------------------------------------------------------------
1372 * Functions for querying mount options/arguments from filesystems.
1373 */
1374
1375/*
1376 * Check that no unknown options are given
1377 */
1378int
1379vfs_filteropt(struct vfsoptlist *opts, const char **legal)
1380{
1381	struct vfsopt *opt;
1382	const char **t, *p;
1383
1384
1385	TAILQ_FOREACH(opt, opts, link) {
1386		p = opt->name;
1387		if (p[0] == 'n' && p[1] == 'o')
1388			p += 2;
1389		for(t = global_opts; *t != NULL; t++)
1390			if (!strcmp(*t, p))
1391				break;
1392		if (*t != NULL)
1393			continue;
1394		for(t = legal; *t != NULL; t++)
1395			if (!strcmp(*t, p))
1396				break;
1397		if (*t != NULL)
1398			continue;
1399		printf("mount option <%s> is unknown\n", p);
1400		return (EINVAL);
1401	}
1402	return (0);
1403}
1404
1405/*
1406 * Get a mount option by its name.
1407 *
1408 * Return 0 if the option was found, ENOENT otherwise.
1409 * If len is non-NULL it will be filled with the length
1410 * of the option. If buf is non-NULL, it will be filled
1411 * with the address of the option.
1412 */
1413int
1414vfs_getopt(opts, name, buf, len)
1415	struct vfsoptlist *opts;
1416	const char *name;
1417	void **buf;
1418	int *len;
1419{
1420	struct vfsopt *opt;
1421
1422	KASSERT(opts != NULL, ("vfs_getopt: caller passed 'opts' as NULL"));
1423
1424	TAILQ_FOREACH(opt, opts, link) {
1425		if (strcmp(name, opt->name) == 0) {
1426			if (len != NULL)
1427				*len = opt->len;
1428			if (buf != NULL)
1429				*buf = opt->value;
1430			return (0);
1431		}
1432	}
1433	return (ENOENT);
1434}
1435
1436char *
1437vfs_getopts(struct vfsoptlist *opts, const char *name, int *error)
1438{
1439	struct vfsopt *opt;
1440
1441	*error = 0;
1442	TAILQ_FOREACH(opt, opts, link) {
1443		if (strcmp(name, opt->name) != 0)
1444			continue;
1445		if (((char *)opt->value)[opt->len - 1] != '\0') {
1446			*error = EINVAL;
1447			return (NULL);
1448		}
1449		return (opt->value);
1450	}
1451	return (NULL);
1452}
1453
1454int
1455vfs_flagopt(struct vfsoptlist *opts, const char *name, u_int *w, u_int val)
1456{
1457	struct vfsopt *opt;
1458
1459	TAILQ_FOREACH(opt, opts, link) {
1460		if (strcmp(name, opt->name) == 0) {
1461			if (w != NULL)
1462				*w |= val;
1463			return (1);
1464		}
1465	}
1466	if (w != NULL)
1467		*w &= ~val;
1468	return (0);
1469}
1470
1471int
1472vfs_scanopt(struct vfsoptlist *opts, const char *name, const char *fmt, ...)
1473{
1474	va_list ap;
1475	struct vfsopt *opt;
1476	int ret;
1477
1478	KASSERT(opts != NULL, ("vfs_getopt: caller passed 'opts' as NULL"));
1479
1480	TAILQ_FOREACH(opt, opts, link) {
1481		if (strcmp(name, opt->name) != 0)
1482			continue;
1483		if (((char *)opt->value)[opt->len - 1] != '\0')
1484			return (0);
1485		va_start(ap, fmt);
1486		ret = vsscanf(opt->value, fmt, ap);
1487		va_end(ap);
1488		return (ret);
1489	}
1490	return (0);
1491}
1492
1493/*
1494 * Find and copy a mount option.
1495 *
1496 * The size of the buffer has to be specified
1497 * in len, if it is not the same length as the
1498 * mount option, EINVAL is returned.
1499 * Returns ENOENT if the option is not found.
1500 */
1501int
1502vfs_copyopt(opts, name, dest, len)
1503	struct vfsoptlist *opts;
1504	const char *name;
1505	void *dest;
1506	int len;
1507{
1508	struct vfsopt *opt;
1509
1510	KASSERT(opts != NULL, ("vfs_copyopt: caller passed 'opts' as NULL"));
1511
1512	TAILQ_FOREACH(opt, opts, link) {
1513		if (strcmp(name, opt->name) == 0) {
1514			if (len != opt->len)
1515				return (EINVAL);
1516			bcopy(opt->value, dest, opt->len);
1517			return (0);
1518		}
1519	}
1520	return (ENOENT);
1521}
1522
1523/*
1524 * This is a helper function for filesystems to traverse their
1525 * vnodes.  See MNT_VNODE_FOREACH() in sys/mount.h
1526 */
1527
1528struct vnode *
1529__mnt_vnode_next(struct vnode **nvp, struct mount *mp)
1530{
1531	struct vnode *vp;
1532
1533	mtx_assert(&mp->mnt_mtx, MA_OWNED);
1534
1535	vp = *nvp;
1536	/* Check if we are done */
1537	if (vp == NULL)
1538		return (NULL);
1539	/* If our next vnode is no longer ours, start over */
1540	if (vp->v_mount != mp)
1541		vp = TAILQ_FIRST(&mp->mnt_nvnodelist);
1542	/* Save pointer to next vnode in list */
1543	if (vp != NULL)
1544		*nvp = TAILQ_NEXT(vp, v_nmntvnodes);
1545	else
1546		*nvp = NULL;
1547	return (vp);
1548}
1549
1550int
1551__vfs_statfs(struct mount *mp, struct statfs *sbp, struct thread *td)
1552{
1553	int error;
1554
1555	error = mp->mnt_op->vfs_statfs(mp, &mp->mnt_stat, td);
1556	if (sbp != &mp->mnt_stat)
1557		*sbp = mp->mnt_stat;
1558	return (error);
1559}
1560
1561void
1562vfs_mountedfrom(struct mount *mp, const char *from)
1563{
1564
1565	bzero(mp->mnt_stat.f_mntfromname, sizeof mp->mnt_stat.f_mntfromname);
1566	strlcpy(mp->mnt_stat.f_mntfromname, from,
1567	    sizeof mp->mnt_stat.f_mntfromname);
1568}
1569
1570/*
1571 * ---------------------------------------------------------------------
1572 * This is the api for building mount args and mounting filesystems from
1573 * inside the kernel.
1574 *
1575 * The API works by accumulation of individual args.  First error is
1576 * latched.
1577 *
1578 * XXX: should be documented in new manpage kernel_mount(9)
1579 */
1580
1581/* A memory allocation which must be freed when we are done */
1582struct mntaarg {
1583	SLIST_ENTRY(mntaarg)	next;
1584};
1585
1586/* The header for the mount arguments */
1587struct mntarg {
1588	struct iovec *v;
1589	int len;
1590	int error;
1591	SLIST_HEAD(, mntaarg)	list;
1592};
1593
1594/*
1595 * Add a boolean argument.
1596 *
1597 * flag is the boolean value.
1598 * name must start with "no".
1599 */
1600struct mntarg *
1601mount_argb(struct mntarg *ma, int flag, const char *name)
1602{
1603
1604	KASSERT(name[0] == 'n' && name[1] == 'o',
1605	    ("mount_argb(...,%s): name must start with 'no'", name));
1606
1607	return (mount_arg(ma, name + (flag ? 2 : 0), NULL, 0));
1608}
1609
1610/*
1611 * Add an argument printf style
1612 */
1613struct mntarg *
1614mount_argf(struct mntarg *ma, const char *name, const char *fmt, ...)
1615{
1616	va_list ap;
1617	struct mntaarg *maa;
1618	struct sbuf *sb;
1619	int len;
1620
1621	if (ma == NULL) {
1622		ma = malloc(sizeof *ma, M_MOUNT, M_WAITOK | M_ZERO);
1623		SLIST_INIT(&ma->list);
1624	}
1625	if (ma->error)
1626		return (ma);
1627
1628	ma->v = realloc(ma->v, sizeof *ma->v * (ma->len + 2),
1629	    M_MOUNT, M_WAITOK);
1630	ma->v[ma->len].iov_base = (void *)(uintptr_t)name;
1631	ma->v[ma->len].iov_len = strlen(name) + 1;
1632	ma->len++;
1633
1634	sb = sbuf_new(NULL, NULL, 0, SBUF_AUTOEXTEND);
1635	va_start(ap, fmt);
1636	sbuf_vprintf(sb, fmt, ap);
1637	va_end(ap);
1638	sbuf_finish(sb);
1639	len = sbuf_len(sb) + 1;
1640	maa = malloc(sizeof *maa + len, M_MOUNT, M_WAITOK | M_ZERO);
1641	SLIST_INSERT_HEAD(&ma->list, maa, next);
1642	bcopy(sbuf_data(sb), maa + 1, len);
1643	sbuf_delete(sb);
1644
1645	ma->v[ma->len].iov_base = maa + 1;
1646	ma->v[ma->len].iov_len = len;
1647	ma->len++;
1648
1649	return (ma);
1650}
1651
1652/*
1653 * Add an argument which is a userland string.
1654 */
1655struct mntarg *
1656mount_argsu(struct mntarg *ma, const char *name, const void *val, int len)
1657{
1658	struct mntaarg *maa;
1659	char *tbuf;
1660
1661	if (val == NULL)
1662		return (ma);
1663	if (ma == NULL) {
1664		ma = malloc(sizeof *ma, M_MOUNT, M_WAITOK | M_ZERO);
1665		SLIST_INIT(&ma->list);
1666	}
1667	if (ma->error)
1668		return (ma);
1669	maa = malloc(sizeof *maa + len, M_MOUNT, M_WAITOK | M_ZERO);
1670	SLIST_INSERT_HEAD(&ma->list, maa, next);
1671	tbuf = (void *)(maa + 1);
1672	ma->error = copyinstr(val, tbuf, len, NULL);
1673	return (mount_arg(ma, name, tbuf, -1));
1674}
1675
1676/*
1677 * Plain argument.
1678 *
1679 * If length is -1, use printf.
1680 */
1681struct mntarg *
1682mount_arg(struct mntarg *ma, const char *name, const void *val, int len)
1683{
1684
1685	if (ma == NULL) {
1686		ma = malloc(sizeof *ma, M_MOUNT, M_WAITOK | M_ZERO);
1687		SLIST_INIT(&ma->list);
1688	}
1689	if (ma->error)
1690		return (ma);
1691
1692	ma->v = realloc(ma->v, sizeof *ma->v * (ma->len + 2),
1693	    M_MOUNT, M_WAITOK);
1694	ma->v[ma->len].iov_base = (void *)(uintptr_t)name;
1695	ma->v[ma->len].iov_len = strlen(name) + 1;
1696	ma->len++;
1697
1698	ma->v[ma->len].iov_base = (void *)(uintptr_t)val;
1699	if (len < 0)
1700		ma->v[ma->len].iov_len = strlen(val) + 1;
1701	else
1702		ma->v[ma->len].iov_len = len;
1703	ma->len++;
1704	return (ma);
1705}
1706
1707/*
1708 * Free a mntarg structure
1709 */
1710static void
1711free_mntarg(struct mntarg *ma)
1712{
1713	struct mntaarg *maa;
1714
1715	while (!SLIST_EMPTY(&ma->list)) {
1716		maa = SLIST_FIRST(&ma->list);
1717		SLIST_REMOVE_HEAD(&ma->list, next);
1718		free(maa, M_MOUNT);
1719	}
1720	free(ma->v, M_MOUNT);
1721	free(ma, M_MOUNT);
1722}
1723
1724/*
1725 * Mount a filesystem
1726 */
1727int
1728kernel_mount(struct mntarg *ma, int flags)
1729{
1730	struct uio auio;
1731	int error;
1732
1733	KASSERT(ma != NULL, ("kernel_mount NULL ma"));
1734	KASSERT(ma->v != NULL, ("kernel_mount NULL ma->v"));
1735	KASSERT(!(ma->len & 1), ("kernel_mount odd ma->len (%d)", ma->len));
1736
1737	auio.uio_iov = ma->v;
1738	auio.uio_iovcnt = ma->len;
1739	auio.uio_segflg = UIO_SYSSPACE;
1740
1741	error = ma->error;
1742	if (!error)
1743		error = vfs_donmount(curthread, flags, &auio);
1744	free_mntarg(ma);
1745	return (error);
1746}
1747
1748/*
1749 * A printflike function to mount a filesystem.
1750 */
1751int
1752kernel_vmount(int flags, ...)
1753{
1754	struct mntarg *ma = NULL;
1755	va_list ap;
1756	const char *cp;
1757	const void *vp;
1758	int error;
1759
1760	va_start(ap, flags);
1761	for (;;) {
1762		cp = va_arg(ap, const char *);
1763		if (cp == NULL)
1764			break;
1765		vp = va_arg(ap, const void *);
1766		ma = mount_arg(ma, cp, vp, -1);
1767	}
1768	va_end(ap);
1769
1770	error = kernel_mount(ma, flags);
1771	return (error);
1772}
1773